An active queue management algorithm for congestion detection and control in dynamic network topologies

One of the main objectives of Active Queue Management (AQM) mechanisms is to provide low delay and low packet loss in best-effort service networks. Nevertheless, AQM algorithms exhibit some weaknesses in the process of congestion detection and congestion whenever we have a dynamic network topology. The paper proposes a new AQM algorithm termed as Adaptive AQM (AAQM) to attain both fairness and congestion control in heterogeneous networks. The performance of AAQM is compared to Blue, Drop-tail, RED and REM using both Reno and Vegas version of TCP. The results based on a large series of simulations with ns (network simulator) indicate that AAQM using Reno is more robust, which means that it is highly responsive to the time-varying state of the network.